Search Issue Tracker
Won't Fix
Votes
0
Found in
2020.1.0a15
Issue ID
1201735
Regression
No
ArticulationBody setup flies around in play mode with no forces applied when there is a falling rigidbody
ArticulationBody setup flies around in play mode with no forces applied when there is a falling rigidbody
To reproduce:
1. Open attached project
2. Enter play mode
Expected result: The capsule/cube body falls to the ground and rests
Actual result: After hitting the ground, he convulsely flies off into the distance
Reproduced in: 2020.1.0a15 (feature introduced in this version)
Note: This only happens if there is a rigidbody cube falling somewhere in the scene (they do not need to collide)
Note2: This only happens if the articulation body has sufficiently complex set up. Disabling any of the 4 connected Articulation Bodies stop the bug from reproducing
Note3: Gif attached
Note4: Capsule (1) has Twist set to Limited. Switching it to Free also makes the bug dissapear
WORKAROUND: One of the limbs has Spherical joint with Twist limited to a [0, 0] range. It creates a bad condition inside PhysX, so increasing the twist range is recommended. Also, I noticed that removing non-uniform scale helped too.
Add comment
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- Particle System Renderer icon is missing in Animation window > Add Property context menu
- [Android] Legacy, TMP and UI Toolkit Input Field language toggle to some non-latin alphabet languages does not work with external keyboard
- Crash on keywords::LocalKeywordState::operator when calling ShaderUtil.GetPasskeywords with ShaderType.RayTracing set as the third parameter
- Inconsistent “Remove property” behavior when removing child properties in Animator window
- Scene View Camera overlay preview goes out of Scene View bounds when the Scene View is resized vertically
Resolution Note:
This is related to a problem in PhysX where a certain type of a spherical joint would cause an ill condition for the algorithm used. It's expected to be addressed in PhysX 5, but we'd only be able to reliably tell when it's out. Closing for now.